Skip to content

Large Language Model (LLM) + RAG knowledge assistant ​

Status: planned. Purpose: help people use existing knowledge more efficiently.

Combine domain databases, large language models and retrieval-augmented generation to find experiment data and similar proposals, then produce domain-specific reports. The pages and workflow below are intended capabilities, not a complete available operating interface.

Expected page content ​

AreaIntended content
Research questionGoals, material system, constraints and required output
Search scopeAuthorized experiment, simulation and literature records and database versions
Similar recordsConditions, results, differences and verifiable sources
Proposal reportEvidence, candidates, conditions, citations and open questions

Target workflow ​

  1. Describe the research question, goal and constraints; choose authorized data.
  2. Retrieve related records and proposals; check sources and applicability.
  3. Generate a report grounded in retrieved evidence, distinguishing facts from recommendations.
  4. Have a researcher review references, methods and conditions before selecting a plan.
  5. Record actual experiments and results in the database, preserving proposal and source relationships.

Use of data and results ​

Retain source, version and original location. Clearly identify missing evidence. Researchers must review reports; generated text is not a validated experimental fact.

Difference from the ML materials agent ​

LLM + RAG helps understand and reuse existing knowledge. The ML materials agent learns relationships in structured data for prediction, screening and inverse design. Both require clear sources and access permissions.

Matching learning resource ​

Built for experiment data workflows.